In contemporary sensor based measurement systems it is often useful to measure not only a behavior of sensors within some expected bounds but also to measure behavior outside those bounds. For instance, in an automotive application various sensors positioned throughout the vehicle are employed to monitor engine operating behavior. These sensors can include sensors for measuring mass air flow, combustion knock, temperature, and exhaust gas oxygen concentration to name a few. Ordinarily these sensors may be expected to provide an output voltage of zero volts to five volts over their full sensing range. Furthermore, the rate of change of the output voltage typically has a relatively slow response. For example, a typical mass air flow sensor is limited to a 33 Hz signal bandwidth, and an exhaust gas oxygen sensor may be limited to a 10 Hz signal bandwidth. As a result of this relatively low signal bandwidth a controller which receives these signals often employs a relatively low signal bandwidth low pass filter to the provided sensor signals to discriminate against noise unassociated with the normal behaviour these sensors are intended to measure.
Unfortunately, since these sensors are electrically connected from their fixed position in the vehicle to a centralized controller, and the vehicular environment is particularly harsh especially in terms of vibration, it is not uncommon that the electrical connection between the sensor circuit and the controller becomes intermittent. This intermittent condition is usually visible as large and fast fluctuations of the sensor's output voltage signal. Since the controller employs a relatively low signal bandwidth low pass filter to the provided sensor signals to discriminate against noise, these intermittent conditions are not interpretable by the controller. If the low signal bandwidth low pass filters are eliminated from the sensory signals the controller can detect this intermittent behavior as well as behavior bounded within the expected bounds but the cost in terms of controller processing time would be significant.
What is needed is a structure for monitoring and analyzing multiple sensory signals for recognizing intermittent or unusual behavior without using excessive controller processing time.